Second Order Step by Step Sliding mode Observer for Fault Estimation in a Class of Nonlinear Fractional Order Systems Seyed Mohammad Moein Mousavi Student Electrical and Computer Engineering Department Tarbiat Modares University Tehran, Iran Email: moein_mousavi@modares.ac.ir Amin Ramezani Professor Electrical and Computer Engineering Department Tarbiat Modares University Tehran, Iran Email: ramezani@modares.ac.ir Abstract This paper considers fault estimation in nonlinear fractional order systems in observer form. For this aim, a step by step second order sliding mode observer is used. By means of a fractional inequality, the stability of the observer estimation errors is analyzed and some conditions are introduced to guarantee finite time convergence of estimation errors. Finally, in a numerical example, effectiveness of this observer is demonstrated. Keyword:sliding mode observer, fault estimation, fractional order system 1 INTRODUCTION Fractional calculus as a generalization of classic calculus is a mathematic tool which is recently being used in control engineering. Among recent years it is proven that some systems are modeled more accurate using fractional order models, compare to integer order ones [2]. Also fault diagnosis has been always an important topic in the industry [4]. Faults can cause failure and damage in physical systems if they are not detected in appropriate time. Fault detection methods are divided into two main types: data based and model based methods. Indeed, model based fault detection methods have been extended during last three decades. In these methods, sensor and actuator faults are detected through the relations between accessible signals. the most popular model based methods are: parameter estimation, observer design [5] and parity space. After a fault is detected, it is important to estimate its shape and domain as an unknown input. Since some systems are modeled by fractional models, using fractional order observers for fault estimation is an important issue. In [6], observability of the states in nonlinear fractional order systems is discussed and using a first order sliding mode observer, fault is estimated as an unknown input in such system. But as the observer is a first order one, the chattering problem exists. Sliding mode observer for fault estimation is discussed in [13] and unknown input observer for fractional order system is considered in [10].in [12] state estimation in a nonlinear fractional order system with uncertain parameters using sliding mode observers is taken into account. sliding mode controller design for a nonlinear fractional order system for disturbance rejection is discussed in [9]. Also in [7,8,11] the structure of a second order sliding mode observer for a nonlinear integer order system is introduced for fault estimation.in [14] a second order sliding mode observer is used for fault estimation in a linear fractional order system. Hence to the best of our knowledge, fault estimation in nonlinear fractional order systems using second order sliding mode observer (which is less affected by the chattering problem) is not discussed in the literature. However, an unknown input observer for this type of systems is designed in [15] which does not consider the unknown input as a state in the observer structure and it is going to be discussed is this paper. In the following parts, some preliminaries on fractional calculus is introduced in section 2. The observer structure and conditions for stability of the state estimation error is introduced in section 3.in section 4 the effectiveness of the proposed observer is discussed by a numerical example and simulation. and finally in section 5 we have the conclusion part. 2 PRELIMINARIES Let be the space of continuous functions on and we mean by the space of real-valued functions with continuous derivatives up to order such that and is the i-th derivative of. 2.1 fractional calculus According to [1] there are three main definitions of fractional order derivatives: - The Riemann Liouville fractional derivative of order of : )1(
- Caputo's derivative of order of : and are known positive constants. the observer structure is considered as: )2( - Grunwald-Letnikov definition: )3( Where denotes the integer part of and is Euler s gamma function. The drawback of the first definition is that the initial conditions are in terms of the variable s fractional order derivatives. having said that, Caputo's definition of fractional derivative needs the initial conditions of the main function and not its fractional derivatives. hence, in engineering usages caputo definition is commonly applied. for the simplicity in the rest of this paper this notation is used: 2.2 Fractional inequality This lemma is proven in [3] which will help us with stability analyze in fractional order systems: Lemma1: Let : Where constant matrix 3 OBSERVER DESIGN )4( be a differentiable vector. For any )5( is a symmetric positive definite and 3.1 Observer structure Let us consider the system in this observable form: )6( Considering fault as a state in the observer structure causes more complexity and the more number of gains to be chosen. This will make the design harder but the estimations will be more accurate. Estimation errors are defined as: With and is the output and the only accessible signal. We have: is a small positive scalar and the observer gains are all positive coefficients. 3.2 Stability analyze Step1: using (6) and (8), error dynamics are obtained as: )8( )9( )10( Where and is the state vector, and are Lipchitz functions and all the states are assumed to be bounded: )7( Using lemma 1: )11( And for all : If the following equations hold, dynamic would be stable and it will converge to zero in bounded time. So and its fractional derivatives would become zero. Considering (10), will equal to.
Step 2: and: )12( 4 NUMERICAL SIMULATION Consider a nonlinear fractional order system as: )13( )14( Similarly, with a suitable choice of observer gains, its derivatives would be 0 and will equal to. Step n: and and denotes the fault to be estimated, observer structure is: And the )15( )16( )17( Again, using lemma1 we have: )18( Because of a similar reason as it is assumed that all the states, fault and their derivatives are bounded, there exists some values for observer gains such that dynamic of becomes stable so converges to and again, if the observer gains are chosen big enough, all the estimations converge to states. Hence the fault signal will be reconstructed. as the output of the system is the only accessible signal and it is the observer input. The initial conditions of the observer parameters are zero and,,.observer gains are,.it should be noted that observer gains are chosen by trial and error. If these gains are chosen small then the estimation errors will not converge to zero, also if we choose big values for the gains, the chattering problem will appear. Therefore, a future work can be optimum choice of these gains. Figure 1. estimation of Figure 2. estimation error of
Figure 3. estimation of Figure 5. estimation of Figure 4. estimation error of Figure 6. estimation error of Simulation Is performed using Grünwald Letkinov's definition of the fractional derivative. State estimations and their errors are depicted in Figures.1-6. Also Figures.7-8 represent the fault signal and its estimation and the estimation error. It is clear that the fault signal is reconstructed and the estimation errors converge to zero. As a second order sliding mode observer is adopted here, it is less affected by the chattering problem compare to first order case. Hence this simulation endorses the effectiveness of the proposed observer. Figure 7. estimation of
5 CONCLUSION Figure 8. estimation error of In this paper a second order step by step sliding mode observer was adopted to estimate states and fault as an unknown input in a nonlinear fractional order system. Using a fractional inequality, conditions for stability and finite time convergence of estimation errors were introduced. a numerical example of a commensurate 3-dimensional nonlinear fractional order system was represented to illustrate the effectiveness of the proposed observer. Since there is no specific method to select observer gains, a future work on this topic can be optimum choice of gains in order to minimize the estimation error. REFERENCES [1] Podlubny I. Fractional differential equations: an introduction to fractional derivatives, fractional differential equations, to methods of their solution and some of their applications. Academic press; 1999. p.198 [2] A.Aribi, C.Farges, M.Aoun, P.Melchior, S.Najar, M.Abdelkrim, Fault detection based on fractional order models: Application to diagnosis of thermal systems, Communications in Nonlinear Science and Numerical Simulation Volume 19, Issue 10, October 2014, Pages 3679 3693. [3] N.Aguila-Camacho,M.A.Duarte-Mermoud,J.A.Gallegos, Lyapunov functions for fractional order systems, Communications in Nonlinear Science and Numerical Simulation Volume 19, Issue 9, September 2014, Pages 2951 2957 [4] M. Blanke, M. Kinnaert, J. Lunze, M. Staroswiecki, Diagnosis and Fault-Telorant Control, Springer Verlag, Heidelberg, 2016 [5] J.Zhang, Evaluation of Observer Structures with Application to Fault Detection, A Thesis for the Degree of Master of Science in Electrical Engineering, Northeastern University Boston, Massachusetts,August 10, 2009 [6] N.Djeghali,S.Djennoune,M.Bettayeb,M.Ghanes,J.Barbot, Observation and sliding mode observer for nonlinear fractional-order system with unknown input, ISA Transactions Volume 63, July 2016, Pages 1 10 [7] A.LEVANT, Higher-order sliding modes, differentiation and output-feedback control,international Journal of Control, 2003, VOL. 76, NOS 9/10, 924 941 [8] T. FLOQUET,J. P. BARBOT, Super twisting algorithm-based step-bystep sliding mode observers for nonlinear systems with unknown inputs, International Journal of Systems Science Vol. 38, No. 10, October 2007, 803 815 [9] S.Pashaei,M.Badamchizadeh, A new fractional-order sliding mode controller via a nonlinear disturbance observer for a class of dynamical system with mismatched disturbances,isa Transactions Volume 63, July 2016, Pages 39 48 [10] I.N Doye,M.Darouach,H.Voos,M.Zasadzinski, Design of unknown input fractional-order observers for fractional-order systems, Int. J. Appl. Math. Comput. Sci., 2013, Vol. 23, No. 3, 491 500 [11] Hu, Z., Zhao, G., Zhang, L. et al., Fault Estimation for Nonlinear Dynamic System Based on the Second-Order Sliding Mode Observer,Circuits,Systems,and Signal Processing January2016, Volume 35, Issue 1, pp 101 115 [12] F.Zhong,H.Li,S.Zhong, State estimation based on fractional order sliding mode observer method for a class of uncertain fractional-order nonlinear systems, Signal Processing Volume 127, October 2016, Pages 168 184 [13] C.Edwards,S.K.Spurgeon, R.J. Patton, Sliding mode observers for fault detection and isolation, Automatica 36 (2000) 541}553 [14] A.Pisano,E.Usai,M.Rapaic,Z.Jelicic, Second-order sliding mode approaches to disturbance estimation and fault detection in fractional-order systems,preprints of the 18th IFAC World Congress Milano (Italy) August 28 - September 2, 2011 [15] U.M.Al-Saggaf,M.Bettayeb,S.Djennoune, Super-Twisting Algorithm- Based Sliding-Mode Observer for Synchronization of Nonlinear Incommensurate Fractional-Order Chaotic Systems Subject to Unknown Inputs, Arab J Sci Eng (2017)